River landform dynamics detection and responses to morphology change in the rivers of North Luzon, the Philippines

Li, Qing (2023) River landform dynamics detection and responses to morphology change in the rivers of North Luzon, the Philippines. PhD thesis, University of Glasgow.

Abstract

River morphology detection has been improved considerably with the application of remote sensing and developments in computer science. However, applications that extract landforms within the active river channel remain limited, and there is a lack of studies from tropical regions. This thesis developed and then applied a workflow employing Sentinel-2 imagery for seasonal and annual river landform classification. Image downscaling approaches were investigated, and the performance of object-based image segmentation was assessed. The area to point regression kriging (ATPRK) approach was chosen to downscale coarser 20 m resolution Sentinel-2 bands to finer 10 m resolution bands. All features were set or processed at 10 m resolution before applying support vector machine (SVM) classification. To improve machine learning classification accuracy, Sentinel-2 acquisitions across one year, which incorporates multiple seasons, should be used. For rivers with different hydrological or geology settings, the thesis considered collecting river specific ground truth data to build a training model to avoid underfitting of models from other hydrological/geological settings. Applying the workflow, three landforms (water, unvegetated bars and vegetated bars) were classified within the active channel of the Bislak, Laoag, Abra and Cagayan Rivers, north Luzon, the Philippines, between 2016 to 2021, respectively. The spatial-temporal river landform datasets enabled the quantitative analysis of the river morphology changes. Water and unvegetated bars showed clear seasonal dynamics in all four rivers, whilst vegetated bars only showed seasonality in the rivers located in the northwest Luzon (the Bislak, Laoag and Abra Rivers). This thesis employed correlated coefficients to investigate the longitudinal correlation between river landforms and active width. It was found that vegetated bar areas always have strong significant correlations (≥0.67) with the active widths in all four rivers, whilst correlation coefficients between vegetated bar areas and active widths in the wet season are higher than that in the dry season. Ensemble empirical mode decomposition (EEMD) was applied to detect landform periodicity; this method indicated that water and vegetated bars commonly showed synchronised fluctuations with precipitation, while unvegetated bars had an anti-phase oscillation with precipitation. In the case of EEMD, deviations from periodic consistency in river pattern may reflect the influence of extreme events and/or human disturbance. Coefficient of variation (COV) was then used to evaluate the stability of the landforms; results suggested that the interplay of faults, elevation, confinement and tributary locations impacted landform stability. Finally, tributary inflow impacts on the mainstem river were investigated for eight tributaries of the lowland Cagayan River, also on Luzon Island. Longitudinal variations in channel morphology and stability, and temporal changes in landform frequency, using Simpson’s diversity index and COV, showed downstream widening associated with tributaries that was controlled by water discharge, with a secondary sediment flux effect. Overall, this thesis provided a novel example of combining remote sensing and GIS science, computing science, statistical science, and river morphology science to study the earth surface processes synthetically and quantitatively within river active channels in the tropical north Luzon, the Philippines. This work demonstrated how the fusion of techniques from these disciplines can be used to detect and analyse river landform changes, with potential applications for river management and restoration.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Supported by funding from China Scholarship Council and University of Glasgow.
• Subjects: G Geography. Anthropology. Recreation > GB Physical geography
• Colleges/Schools: College of Science and Engineering > School of Geographical and Earth Sciences
• Funder's Name: China Scholarship Council
• Supervisor's Name: Williams, Professor Richard and Barrett, Dr. Brian
• Date of Award: 2023
• Depositing User: Theses Team
• Unique ID: glathesis:2023-83661
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Jun 2023 12:51
• Last Modified: 10 Apr 2024 14:10
• Thesis DOI: 10.5525/gla.thesis.83661
• URI: https://theses.gla.ac.uk/id/eprint/83661

Related URLs

• Data DOI